Skip to content

10 · Capstone — Full Product Strategy + Org Design

This is the deliverable the whole path has been building toward: a three-year product strategy for ListUp, and the organisation that can execute it, written as one document a board would actually accept.

The reason strategy and org design appear in the same capstone is that they are the same decision seen from two sides. A strategy that no structure can execute is a wish; an org chart with no strategy behind it is decoration. Most companies produce both artefacts and never reconcile them, which is why so many strategies survive exactly until the first planning cycle.

Below is a complete worked deliverable. Read it as a model, then produce the same nine sections for your own product.

What the deliverable contains

§ Section Answers Length
1 Diagnosis What is actually going on, including the uncomfortable part 1 page
2 Guiding policy Where we play and how we win Half page
3 Coherent actions The three or four things we will do, and what we will not 1 page
4 Financial model What this produces, top-down and bottom-up 1 page
5 Org design The structure that can execute it 1 page
6 Hiring and bench Who we need and when Half page
7 Operating cadence How decisions get made through the year Half page
8 Risks and kill criteria What would make this wrong, and when we stop Half page
9 One-page strategy The whole thing, memorable, on one page 1 page

§ 1 · Diagnosis

Where ListUp is. $52.4M ARR growing 37.9%. Three lines: Publish ($27.2M, 19%), Decide ($14.7M, 71%), Console ($10.5M, 62%). 6,150 direct accounts contributing $18.2M and 214 agencies contributing $34.2M. 310 employees, 76 in R&D. Net revenue retention 112% blended, 116.7% direct. Rule of 40 at 27.6 — below the threshold and below plan.

Four facts that constitute the actual problem:

Fact Evidence Why it matters
The revenue engine is the channel we understand least Agencies are 65.3% of ARR from 214 relationships; direct is 34.7% from 6,150 accounts Concentration risk, and the roadmap is written for the smaller half
Half the direct base is decaying 3,210 Publish-only accounts, 52.2%, at 97% NRR The largest identified opportunity in the company, unaddressed
The growth is carried by one young line Decide at 71% on $14.7M; Publish at 19% on $27.2M Momentum depends on a line whose model has a known error rate
Efficiency is below par while growth is above it Rule of 40 at 27.6; R&D at 28.3% of revenue The next round or the next board will price this

The uncomfortable part, stated plainly. ListUp is a publishing tool that happens to sell intelligence, being paid mostly by agencies, while building mostly for direct sellers. Every one of those clauses is defensible on its own and together they describe a company whose investment does not match its revenue.

Competitive position. Channelry is larger, better funded, and better at channel breadth — more marketplaces, faster integrations. It has no equivalent of Decide's 4.1M labelled price changes and no agency workflow worth the name. It will reach parity on agency workflow in roughly 18 months if nothing changes, and it will not reach parity on the pricing dataset within the plan period, because that asset compounds with usage.

The diagnosis in one sentence: ListUp's durable advantage is a proprietary pricing dataset sold through an agency channel it under-serves, while it invests as though it were a self-serve publishing tool.

§ 2 · Guiding policy

Where we play. Multi-channel commerce operations for sellers doing $2M–$200M of online GMV, reached primarily through the agencies and aggregators that manage them.

How we win. A pricing and merchandising intelligence layer that improves with every catalogue it sees, delivered inside the workflow the agency already runs. Publishing is the entry point and the data source; it is not the product we win on.

What we are explicitly not doing. Not competing on marketplace breadth with Channelry. Not building a storefront or a commerce platform. Not pursuing enterprise direct sales below $500k ACV. Not launching a fourth product line inside the plan period.

The fourth exclusion is the one that will be tested most often, by the most senior people, and it is the reason to write it down now.

§ 3 · Coherent actions

# Action Why it follows from the diagnosis Owner
1 Make Decide the default, not the upsell — attach it to every Publish account with a 60-day zero-friction trial, and move the paywall to a value boundary 3,210 accounts at 97% NRR; attach moves them to 116% Decide group
2 Build the agency operating system — multi-client workflow, approvals, white-label reporting, and an agency API 65.3% of revenue, served by a product built for single sellers Console group
3 Compound the data asset deliberately — instrument every accept/reject as a labelled example; per-slice evaluation; the review queue as a training-data engine It is the one advantage Channelry cannot buy Decide + Platform
4 Hold Publish at parity, not at leadership Channel breadth is Channelry's game; parity is enough to keep the entry point Publish group

What we stop: the enterprise direct pipeline below $500k ACV (three in-flight deals released to partners), the mobile app rewrite, and one-off channel adapters requested by single agencies.

Why this set is coherent rather than a list: action 1 creates the usage that produces the labels for action 3, which improves the model that makes action 2's agency proposition defensible, which grows the channel that distributes action 1. Publish is held flat deliberately because it feeds all three and wins none of them. A list of four good initiatives that do not reinforce each other is a budget, not a strategy.

§ 4 · Financial model

Top-down plan, with explicit deceleration:

Year Growth ARR
Now $52,400,000
Year 1 38% $72,312,000
Year 2 32% $95,451,840
Year 3 27% $121,223,837

Three-year CAGR: 32.3%.

Bottom-up by line, at each line's own decelerating rate:

Line Now Year 1 Year 2 Year 3 Rates
Publish $27,200,000 $32,368,000 $37,546,880 $42,803,443 19 / 16 / 14%
Decide $14,700,000 $25,137,000 $38,962,350 $56,105,784 71 / 55 / 44%
Console $10,500,000 $17,010,000 $25,515,000 $35,721,000 62 / 50 / 40%
Total $52,400,000 $74,515,000 $102,024,230 $134,630,227 42.2 / 36.9 / 32.0%

The bottom-up is 11.1% above the plan, and the plan is the number we commit to. The gap of $13,406,390 is the deceleration we believe in but cannot yet evidence, and the discipline is to show both, name the gap, and commit to the lower one. A leader who takes the bottom-up number to a board gets one good quarter and then spends two years explaining a miss.

Year-3 mix at the plan number:

Line Year 3 ARR Share Share now
Publish $38,541,104 31.8% 51.9%
Decide $50,518,807 41.7% 28.1%
Console $32,163,926 26.5% 20.0%

Channel mix, the strategy's real test:

Channel Now Share Year 3 Share Implied CAGR
Direct $18,187,200 34.7% $52,126,250 43.0% 42.0%
Agency $34,212,800 65.3% $69,097,587 57.0% 26.4%

Agencies grow from 214 to roughly 320 at an average of $215,930 each, while direct grows faster because action 1 converts the Publish-only base. Concentration falls without the agency business shrinking, which is the only acceptable version of de-risking a channel that funds you.

The single largest identified move, sized: converting 25% of the 3,210 Publish-only accounts to Publish + Decide is 802 accounts × ($3,480 − $1,560) = $1,540,800, or 2.94% of current ARR, and it moves those accounts from 97% to 116% NRR, which compounds every year afterwards. The caveat from Level 4, Module 2 stands: until the matched comparison is run, this is the size of the prize if the relationship is causal.

§ 5 · Org design

Now: three product groups plus Platform, 76 R&D people, 52 engineers. Year 3: the same four groups, with Decide split and one new capability team.

Group Eng now Eng yr 3 Owns Accountable for
Publish 14 15 Publishing pipeline, marketplace adapters Reliability, parity, $38.5M line
Decide — Applied 15 20 Recommendations, thresholds, review queue, auto-apply Attach, accepted-change value, $50.5M line
Decide — Model (within) 12 Model, evaluation, data labelling engine Per-slice accuracy, break-even precision, dataset growth
Console 13 21 Agency workflow, approvals, white-label, agency API Agency NRR, agencies onboarded, $32.2M line
Platform 10 12 Events, API, identity, billing, cost controls Internal SLAs, inference cost per account
Total 52 80

The one structural change that matters is splitting Decide into Applied and Model. Today one group owns both the model and the product surface, and the model work is starved every time the product surface has a deadline — which is every quarter. Splitting them makes the dataset a funded product with its own roadmap, its own metrics, and a platform-style contract with Applied as its named internal customer. It is action 3 expressed as structure, which is the test of whether an action is real.

What this structure costs, stated in advance: two Decide groups will disagree about thresholds, and that argument now escalates rather than being settled in a standup. Mitigation is a written contract — Model owns per-slice accuracy and the evaluation set, Applied owns the threshold and the business outcome, and the break-even precision arithmetic from Level 4, Module 7 arbitrates between them.

Growth in R&D against revenue:

Year Revenue R&D people Engineers R&D cost % of revenue Revenue per R&D head
Now $52,400,000 76 52 $14,820,000 28.3% $689,474
Year 1 $72,312,000 88 60 $17,160,000 23.7% $821,727
Year 2 $95,451,840 102 70 $19,890,000 20.8% $935,802
Year 3 $121,223,837 116 80 $22,620,000 18.7% $1,045,033

R&D falls from 28.3% to 18.7% of revenue without a single reduction, purely because revenue compounds faster than headcount. That is the entire Rule-of-40 recovery, and it is worth stating explicitly to a board, because otherwise someone will propose achieving it by cutting.

§ 6 · Hiring and bench

At 6.5 engineers per PM, 80 engineers needs 12.3 PMs against 8 today.

Role When Why
Director, Decide Q1 yr 1 The split needs a leader before it needs headcount
PM, Decide Model Q2 yr 1 The dataset needs an owner, not a stakeholder
PM, Agency API Q2 yr 1 Action 2's hardest surface
Group PM, Console Q4 yr 1 Console reaches two squads' worth of scope
2 × PM Yr 2 Ratio maintenance
Director, Console Yr 2 Second Director; VP span stays at 4
2 × PM, 1 × Group PM Yr 3 Ratio maintenance

Bench, honestly assessed: two of the four Director-level roles in year 3 have a ready internal successor; two do not. The named development assignments for the two internal candidates start in Q1 of year 1, because the alternative is discovering the gap in the quarter you need it filled.

§ 7 · Operating cadence

The base cadence is the one from Level 4, Module 1. This strategy adds two forums, because two of its actions need a decision-forcing venue that does not currently exist.

Cadence Forum Decision it can force
Monthly Model review — per-slice accuracy, thresholds, inference cost by decile Halt a model release on a slice regression; change the auto-apply threshold
Half-yearly Strategy check against § 8 Stop or double an action, on the criteria rather than on the argument

Everything else — weekly product leadership, monthly product review, quarterly portfolio review and allocation, annual org design — stays as it is. A strategy that requires a new meeting for every action has not been delegated.

§ 8 · Risks and kill criteria

Risk Leading indicator Kill / pivot criterion
Decide attach does not move the Publish-only base Trial-to-paid on the 60-day trial Below 8% by end of Q3 yr 1 → rebuild the offer, not the model
Channelry reaches agency parity early Agency win rate in competitive deals Below 50% for two quarters → re-price or re-scope action 2
Model advantage does not compound Labelled examples per month; per-slice accuracy trend Flat for two quarters → the dataset is not the moat; rewrite § 1
Agency concentration worsens Top 10 agencies as % of ARR Above 30% → direct becomes the priority regardless of efficiency
Inference cost inverts on heavy users Gross margin on AI add-ons, by decile Blended below 75% → meter it
The Decide split fails Threshold disputes escalating to the VP More than 2 per quarter → the contract is wrong, fix it before restructuring
Rule of 40 does not recover Growth plus FCF margin, quarterly Below 30 at end of yr 1 → the deceleration assumption was wrong, not the spend

Every criterion is a number and a date, and each fires automatically. Anything that has to be argued for at the moment of failure will not be argued for.

§ 9 · The one-page strategy

Diagnosis. Our durable advantage is a proprietary pricing dataset sold through an agency channel we under-serve, while we invest as though we were a self-serve publishing tool.

Where we play. Multi-channel sellers doing $2M–$200M GMV, reached through the agencies that manage them.

How we win. Pricing intelligence that improves with every catalogue it sees, delivered inside the agency's own workflow.

What we do. Make Decide the default, not the upsell. Build the agency operating system. Compound the dataset deliberately. Hold Publish at parity.

What we don't do. Channel breadth. Storefronts. Sub-$500k enterprise direct. A fourth product line.

What success looks like in three years. $121M ARR at 32% CAGR, Decide at 41.7% of revenue, agency concentration down from 65.3% to 57% without shrinking, R&D at 18.7% of revenue, and Rule of 40 above 40.

What would prove us wrong. Decide trial-to-paid below 8% by Q3, or the labelled dataset flat for two quarters.

Every person in the org should be able to state the second, third and fourth paragraphs from memory. The measure of whether they can is asking five people at random and writing down the variance in their answers.

How this deliverable is judged

Criterion Fails when Passes when
Diagnosis Describes the market Names the uncomfortable thing about this company
Guiding policy Says what you will do Says what you will not do, specifically
Coherence Four good initiatives Each action makes the next one work
Model One optimistic number Top-down and bottom-up, with the gap named and the lower one committed
Org design Boxes Structure derived from the actions, with its costs stated
Bench Headcount only Named successors and named gaps
Risk A generic list Numbers and dates that fire automatically
One-pager A summary Repeatable from memory by someone who did not write it

How It Actually Works: why compounding rates, not levels, drive every number in this capstone

The Rule-of-40 recovery is pure exponential arithmetic, and it is worth tracing because it is the mechanism the whole capstone quietly relies on. R&D cost grows at roughly the rate of engineer headcount (14.3% CAGR: 52 to 80 over 3 years, (80/52)^(1/3) - 1 ≈ 15.5%), while revenue grows at 32.3% CAGR. Two exponentials with different bases diverge — R&D% = R&D_cost(t) / Revenue(t) = R&D_cost₀(1+r_cost)^t / Revenue₀(1+r_rev)^t = (R&D%)₀ × ((1+r_cost)/(1+r_rev))^t. Since 1.155/1.323 ≈ 0.873 < 1, this ratio shrinks geometrically every year regardless of any efficiency initiative — purely because the denominator's growth rate exceeds the numerator's. This is precisely why the module insists the recovery must be shown as a consequence of the growth/cost-rate gap, not claimed as an achievement of cost discipline: if someone proposes achieving the same 18.7% via cuts, they are solving a problem the compounding math had already solved, at the cost of the growth rate that was doing the real work.

Top-down vs. bottom-up divergence is what a sum of compounding sub-rates always produces relative to a single blended rate, and the gap grows with the dispersion between the sub-rates, not their average. The bottom-up model compounds each line at its own (higher-dispersion) rate — Decide at 71/55/44%, Publish at 19/16/14% — while the top-down applies one smoothed rate (38/32/27%) to the whole. Because compounding is convex in the growth rate (a mix of high and low compounding sub-populations produces a higher sum than the same weighted-average rate applied uniformly, by Jensen's inequality applied to the convex function (1+r)^t), the bottom-up total is mechanically guaranteed to sit at or above a same-weighted-average top-down projection whenever the underlying growth rates genuinely differ — which is exactly the $13.4M, 11.1% gap observed, and exactly why committing to the lower, smoothed number is the structurally conservative choice rather than an arbitrary hedge.

The NRR-driven "single largest identified move" ($1,540,800) is the same compounding-exponent argument from Level 4 Module 2, scaled to the whole capstone. Moving 802 accounts from 97% to 116% NRR doesn't just add $1.54M this year — it changes those accounts' terminal-value exponent from 0.97^t to 1.16^t. Over the 3-year plan horizon that is 1.16³ = 1.56x versus 0.97³ = 0.91x, a 1.7x gap in accumulated value per dollar of starting ARR converted — which is why the strategy treats this move as structurally different in kind from ordinary upsell revenue, and why the matched-comparison caveat (is the NRR lift causal, or do better accounts just buy more?) is flagged as the single biggest uncertainty in the entire financial model: an error in the causal attribution compounds at the same exponential rate as the effect itself.

Why kill criteria are placed on leading indicators (trial-to-paid rate, labelled-examples-per-month) rather than lagging ones (ARR, NRR). A leading indicator changes on the timescale of the underlying mechanism (a trial converts in 60 days); a lagging one, like ARR, only reflects that change after it has compounded through a much longer accumulation period. Because compounding effects are slow to reveal themselves and fast to become expensive once they diverge, a kill criterion set on a lagging metric detects failure only after several compounding periods have already locked in the wrong trajectory — mechanically the same "detection gap" argument as the incident-response module's 47-minute cost, applied to a multi-quarter strategic bet instead of a multi-hour production incident.

Stretch goals

  1. Run the matched comparison that settles the attach causality question: accounts matched on size, tenure and channel count, NRR before and after attach. Rewrite § 4's largest move with the corrected number, and state how much of the $1,540,800 survives.
  2. Build the sensitivity model. Recompute the three-year plan at Decide growing 55/40/30 instead of 71/55/44, and write the org design that version implies. If the structure does not change, one of the two documents is not real.
  3. Write the losing version. Produce the strategy Channelry would write to beat this one, in the same nine sections, then list the three things it exploits and what you would change in response.
  4. Cost the org design properly, including the productivity dip of the Decide split — four to eight weeks — and the recruiting cost of 12 hires at senior product level. Compare the total against the incremental ARR the structure is supposed to unlock.
  5. Write the year-1 falsification review in advance: the exact table you will fill in twelve months from now, with the kill criteria as rows and blank cells for the actuals. Put the date in the calendar before you present the strategy.
  6. Reconcile the strategy with individual objectives. Trace one action from § 3 down to a named PM's quarterly objective, and back up again. Any action that does not survive the round trip is not funded, whatever the document says.
  7. Present it in ten minutes to a hostile audience using Level 4, Module 5's structure — bottom line, evidence, risk, ask — and have someone play the board member whose only question is why growth decelerates from 38% to 27% while headcount rises 53%.